Applying deep learning algorithms for non-invasive estimation of carotenoid content in foot muscle of different colors in Pacific abalone

Carotenoids are vital pigments influencing both the coloration and health of aquatic organisms, particularly in species such as the Pacific abalone (Haliotis discus hannai). In this study, we identified the major carotenoids in abalone foot muscle using targeted metabolomics. Through differential metabolite analysis, we selected metabolites that met the following criteria: p-value <0.05, variable importance in projection (VIP) score ≥ 1, and fold change (FC) ≥ 2 or FC ≤ 0.5. The results showed that zeaxanthin had the highest content among all foot muscle colors, with the most significant p-value of 0.0079. Thus, we confirmed that zeaxanthin is the predominant carotenoid contributing to the distinct coloration of the foot muscle. We then used a deep learning model to predict carotenoid content based on color measurements in the CIELAB color space, defined by the Commission Internationale de l'Eclairage (CIE), which includes three dimensions: lightness (L*), redness-greenness (a*), and yellowness-blueness (b*). Performance evaluation of 344 abalone samples showed that the Long Short-Term Memory (LSTM) model provided the best prediction results, with a root mean square error (RMSE) of 6.692 and a coefficient of determination (R²) of 0.415. Furthermore, we developed the Color-Based Carotenoid Estimation Suite (CCES). This software features a user-friendly graphical interface, enabling users to input colorimetric data, train models, and predict carotenoid content. Compared to traditional methods, CCES offers non-destructive, rapid carotenoid estimation, improving efficiency by 450 times and reducing costs by 47 to 77 times. This method provides an efficient and scalable tool for aquaculture breeding and quality control, with applications extending beyond abalone to other aquatic and terrestrial species.